Abstract
With the development of Geographic Information Systems (GIS), crime mapping becomes an effective approach to investigate the spatial pattern of crime in a defined area. Understanding the relationship between crime and its surrounding environment can reveal possible strategies that can reduce crime in a neighbourhood. The relationship between vegetation density and crime has been under debate for a long time. This research is conducted to investigate the impacts of tree coverage on property crime in the City of Vancouver. High spatial resolution airborne LiDAR data collected in 2013 was used for the extraction of tree covered area for cross-sectional analysis. The independent variables were put into Ordinary Least-Squares (OLS) regression, Spatial Lag regression, and Geographically Weighted Regression (GWR) models to examine their influences on property crime rates. According to the results, the cross-sectional analysis demonstrated statistical evidences that property crime rates had negative correlations with tree coverage, with greater influences occurred around Downtown Vancouver.
Introduction
The relationship between urban vegetation and crime has been a subject of academic debate for decades. The "concealment hypothesis" suggests that dense foliage provides cover for criminals to hide or stash stolen goods. Conversely, the "surveillance and restoration hypothesis" argues that green spaces attract more people outdoors (increasing "eyes on the street") and provide a mentally restorative effect that reduces aggression. While many studies have explored this in the U.S., there is a lack of high-resolution spatial analysis in Canadian contexts. This paper addresses that gap by investigating how tree canopy coverage influences property crimes—specifically theft, break and enter (BNE), and mischief—across Vancouver’s diverse neighborhoods.
Methods and Data
The study integrated several high-quality datasets from 2013, including airborne LiDAR (168 tiles at 12 points/m²), geocoded property crime data from the Vancouver Police Department’s "GeoDash" portal, and Canadian Census data for demographic variables such as population density, income, and unemployment.
The researchers employed a multi-stage spatial analysis workflow:
- Feature Extraction: Airborne LiDAR data was used to extract precise tree crown polygons. This allowed for a much higher spatial resolution (2m) than traditional satellite imagery, accounting for trees near buildings and along narrow streets.
- Statistical Modeling: Three types of models were compared to move beyond simple correlations:
- Ordinary Least-Squares (OLS): The baseline linear regression.
- Spatial Lag Model: This accounts for "spatial autocorrelation," recognizing that crime in one neighborhood often spills over into adjacent areas.
- Geographically Weighted Regression (GWR): This sophisticated approach allows the influence of trees to vary by location, revealing local variations in the relationship between greenery and crime.
Results and Discussion
The LiDAR-based tree extraction achieved an overall accuracy of 98.4%. Statistical analysis revealed a significant inverse relationship across the city: tree coverage had a statistically significant negative correlation (p < 0.01) with total property crime, theft, and BNE.
As shown in the regression results, the Spatial Lag model significantly outperformed the OLS model (e.g., for property crime, a Pseudo R² of 0.45 vs. 0.20), proving that crime analysis cannot ignore the "neighbor effect." BNE showed a smaller magnitude of influence from trees compared to general theft, likely because BNE is more heavily influenced by building security and family income levels.
The GWR analysis provided deeper insights into spatial variations. The negative correlation between trees and crime was strongest in high-density areas like Downtown Vancouver and the Strathcona neighborhood, suggesting that adding trees in these areas has the highest marginal benefit for crime reduction.
Interestingly, exceptions were noted in Stanley Park and parts of Kensington-Cedar Cottage, where positive correlations were observed. The authors argue this is not because trees cause crime, but because other factors dominate: Stanley Park is a high-traffic tourist hub vulnerable to theft, and Kensington-Cedar Cottage faces high "social disorganization" (poverty and residential mobility) that outweighs environmental benefits.
Conclusion and Future Work
The results provide empirical support for Vancouver’s "Greenest City Action Plan," suggesting that urban forestry and public safety goals are mutually reinforcing. Future work should address limitations such as the exclusion of violent crime data due to privacy restrictions. Additionally, utilizing LiDAR’s vertical dimension to investigate if high-canopy trees have different effects than low-level shrubs—which might aid concealment—could further refine urban planning strategies.
Summary updated